1. Field of the Invention
The present invention relates to the control of a monolithic photosensitive cell of an image sensor for use in image shooting devices such as, for example, film cameras, camcorders, digital photographic devices, or again cellular phones. More specifically, the present invention relates to a semiconductor-based photosensitive cell.
2. Discussion of the Related Art
FIG. 1 schematically illustrates the circuit of a photosensitive cell PIX, or pixel, of an array of photosensitive cells of an image sensor. With each photosensitive cell of the array are associated a precharge device and a read device. The precharge device is formed of an N-channel MOS transistor M1, interposed between a supply rail VRT and a sense node SN. The gate of precharge transistor M1 is capable of receiving a precharge control signal RST. The read device is formed of the series connection of first and second N-channel MOS transistors M3, M2. The source of first read transistor M2 is connected to an input terminal X of a processing circuit (not shown). The drain of second read transistor M3 is connected to supply rail VRT. The gate of first read transistor M2 is connected to sense node SN. The gate of second read transistor M3 is capable of receiving a read signal READ. The photosensitive cell comprises a photodiode D having its anode connected to a reference voltage source GND and having its cathode connected to node SN via an N-channel charge transfer transistor M4. The gate of transfer transistor M4 is capable of receiving a charge transfer control signal TG. Generally, signals READ, RST, and TG are provided by control circuits not shown in FIG. 1 and may be provided to all the photosensitive cells of a same row of the cell array.
Sense node SN function as a region for storing the charges originating from photodiode D, the apparent capacitance at sense node SN being formed of the capacitances of the sources of transistors M1 and M4, of the input capacitance of transistor M2, as well as on the set of stray capacitances present at node SN. According to an alternative, a specific component, for example, a diode or a capacitor, may be connected to sense node SN to ensure the storage function.
FIG. 2 shows an example of a timing diagram of signals READ, RST, TG and of voltage VSN at node SN for a read cycle of photosensitive cell PIX of FIG. 1. Signals READ, RST, and TG are binary signals varying between high and low signals which may be different for each of the signals.
Between two read cycles of the photosensitive cell, during the carrier integration phase, signal TG is low. Transfer transistor M4 is thus turned off. The lighting causes the forming and the storage of charges at the level of photodiode D. Further, signal RST is high. Precharge transistor M1 is thus on. Voltage VSN is thus substantially equal to voltage VRT.
At a time t0, the array row containing the photosensitive cell to be read is selected by setting signal READ to the high level. The precharge of sense node SN is interrupted by setting signal RST to the low state at time t1, thus turning off precharge transistor M1. Voltage VSN at sense node SN is then set to a precharge level VRST which is slightly lower than voltage VRT due to a coupling with precharge transistor M1. Precharge level VRST is generally disturbed by noise essentially originating from the thermal noise of the channel of precharge transistor M1. This noise is sampled and maintained at sense node SN on turning off of precharge transistor M1. Precharge level VRST is then stored outside of photosensitive cell PIX via read transistors M2, M3.
At time t2, signal TG switches high. Voltage VSN rises from VRST to VRST+VU due to the coupling with transistor M4. Transfer transistor M4 is then on, which enables transferring the charges stored in photodiode D to sense node SN, causing a decrease in voltage VSN down to VRD+VU. Photodiode D is designed so that all the charges stored therein are transferred to sense node SN. Once the charge transfer is over, signal TG switches low at time t3, thus enabling insulating again photodiode D and restarting a cycle of forming and storage of charges resulting from the lighting. By a coupling effect with transistor M4, voltage VSN then decreases to stabilize at a desired signal level VRD, smaller than level VRST, which depends on the number of charges transferred to sense node SN. Wanted signal level VRD is then read via read transistors M2, M3. Like precharge level VRST, desired signal level VRD is especially disturbed by the thermal noise of the channel of precharge transistor M1 which has been sampled and maintained at sense node SN. The subtraction of signals VRD and VRST by the processing circuit enables eliminating the noise of precharge transistor M1 by a double correlated sampling. Once the reading is over, signal RST is set to the high state at time t4 to precharge sense node SN again. Finally, at time t5, signal READ is set to the low state to deselect the photosensitive cell. According to a variation, the switchings between high and low levels of signal TG are performed in ramps.
FIG. 3 illustrates, in a partial simplified cross-section view, a monolithic embodiment of the assembly of photodiode D and of transfer transistor M4 of FIG. 1. These elements are formed in the same active area of a lightly-doped semiconductor substrate 1 of a first conductivity type, for example, type P (P−). This substrate, for example, corresponds to an epitaxial layer on a silicon wafer which forms reference plane GND. The active area is delimited by field insulation areas 2, for example, made of silicon oxide (SiO2), and corresponds to a well 3 of the same conductivity type as underlying substrate 1, but more heavily-doped. Above the surface of well 3 is formed an insulated-gate structure 4 possibly provided with lateral spacers. On either side of gate 4, at the surface of well 3, are located source and drain regions 5 and 6 of the opposite conductivity type, for example, N. Drain region 6, to the right of gate 4, is heavily doped (N+). Source region 5 is formed of a much greater surface area than drain region 6 and forms with underlying well 3 the junction of photodiode D. Gate 4 and drain 6 are solid with metallizations (not shown) which enable putting these regions respectively in contact with transfer control signal TG and the gate of transistor M2 (node SN), respectively. The structure is completed with heavily-doped
P-type regions 8 and 9 (P+). Regions 8 and 9, which underly areas 2, are connected with the reference potential or ground via well 3 and substrate 1. Photodiode D is of so-called pinned or fully depleted photodiode type and comprises, at the surface of its source 5, a shallow P-type region 7 more heavily doped (P+) than well 3. Region 7 is in lateral (vertical) contact with region 8. It is thus permanently maintained at the reference voltage. Photodiode D is called a depleted or pinned photodiode since the voltage of region 5 of the photodiode is, in the absence of lighting, set by the sole dopant concentrations of regions 3, 5, 7.
FIG. 4 schematically illustrates the voltage levels of the different regions of FIG. 2. The curve in stripe-dot lines illustrates the state of the system just after time t2, and the curve in full line illustrates the state of the system just after time t3. Heavily-doped P-type regions 7, 8, and 9 are permanently maintained at the reference potential or ground, for example, 0 V. Just after time t2, region 5 of photodiode D, completely charged, is at a voltage VDC. Transistor M4 is on. Channel region 3 of transistor M4 is at a voltage VT. Region 6 corresponding to node SN is at the level of precharge level VRST+VU due to the coupling with transistor M4. Between times t2 and t3, the charges stored in region 5 are transferred to region 6, causing a decrease in the voltage of region 6 and an increase in the voltage of region 5. After time t3, the charges stored in photodiode D being completely transferred to node SN, photodiode D reaches a so-called quiescent depletion level VD set by the sole characteristics of diode D. Transfer transistor M4 being off, channel region 3 is at 0 V. Region 6 is at desired signal level VRD. Region 5 of photodiode D then forms an empty voltage well which fills up again according to the lighting of the photodiode.
Generally, the high level of transfer control signal TG applied to the gate of transfer transistor M4 is such that the voltage in channel region 3 of transistor M4 is intermediate between depletion level VD and desired signal level VRD increased by voltage VU due to the coupling with transistor M4. To ensure a proper transfer of the charges, it is generally necessary to provide a sufficient margin M between voltage levels VD and VT. As an example, for a supply voltage VRT of 3 V, voltage VD is on the order of 1.5 V and margin M is generally selected to be greater than 0.5 V. Voltage VT thus defines the swing of voltage VSN which substantially corresponds to the difference between voltages VRST+VU and VT.
For increasingly dense technologies with photosensitive cells of small dimensions, it is desired to decrease supply voltage VRT and, generally, the high levels of the transistor control signals.
However, several difficulties arise when supply voltage VRT is decreased. A first difficulty is that the decrease of voltage VRT translates as a decrease in voltage VRST. With the previously-described image sensor controlling method, it can then be difficult, or even impossible, to adjust voltages VD and VT to ensure the proper transfer of the charges of photodiode D to sense node SN while keeping an appropriate swing of voltage VSN and an appropriate swing of the photodiode voltage.
Another difficulty is to ensure for transistor M2 to operate in linear state across the entire swing of voltage VSN to ensure for the voltage at node X to be a linear reproduction of voltage VSN. Transistor M2 is said to be in linear state when the ratio of voltages VX and VSN varies only slightly. According to the level of supply VRT, the linear state of transistor M2 corresponds to a specific range of voltage VSN. With the previously-described image control method, it may be difficult, or even impossible, when VRT is desired to be decreased, to have the range of voltage VSN for which transistor M2 is in linear state correspond to the range of voltage VSN for which a proper charge transfer from photodiode D to sense node SN is obtained.